Pg 79 of "Tensors, Relativity and Cosmology"

In order to construct the geodesic equations which define the curve with a stationary arc length, we may choose the arc length itself as the action integral with zero variation. Using $$s(t)=\int_{t_o} ^t\sqrt{g_{mn}\frac{dx^m}{dt}\frac{dx^n}{dt}}dt \tag{1}$$

we write $$I=\int_{S_A}^{S_B}\sqrt{g_{mn}\frac{dx^m}{ds}\frac{dx^n}{ds}}ds \tag{2}$$

where $I$ is the action integral and is equal to unity along the geodesic line C, where $$ds^2=g_{mn}dx^mdx^n \tag{3}$$

But isn't (2) just equal to 1 since $\frac{ds}{ds}=1$ and can this still be an action integral (functional) if the LHS of (2) is just a constant?

Or was it the author's intention to introduce to a scalar parameter such as $d\lambda$ but accidentally used ds instead?

(The author actually made tons of mistakes in the previous chapters, so I'm just making sure that this is not another error but I could be wrong...)

  • $\begingroup$ Or to put it another way, under what circumstances is (3) true? $\endgroup$
    – puppetsock
    Jun 14, 2019 at 14:34

2 Answers 2


Yes, OP is right. One cannot consistently formulate a variational principle for geodesics with arc length/affine parameter as the independent curve parameter for all virtual curves. Although the square root action is indeed independent of parametrization $\lambda$, the point is that the parameter interval $[\lambda_i,\lambda_f]$ must be common for all virtual curves, so that if $\lambda$ happens to be the arc length/affine parameter for one particular curve, this will not be the case for all neighboring curves. For more details, see e.g. my Phys.SE answer here.

  • 1
    $\begingroup$ The author might (admittedly confusingly) claim that the notation $s$ in eq. (1) denotes arc length while the notation $s$ in eq. (2) denotes an arbitrary parameter. $\endgroup$
    – Qmechanic
    Jun 14, 2019 at 11:49
  • $\begingroup$ And this arbitrary parameter cannot be/is not necessary an affine parameter/arc length parameter? $\endgroup$ Jun 14, 2019 at 13:14
  • $\begingroup$ Not for all virtual curves. $\endgroup$
    – Qmechanic
    Jun 14, 2019 at 19:08

One can choose any parameter one likes to parameterize a curve and the standard trick is to use $s$ itself (sometimes called "proper time"). So the length of a curve is defined to be $$ L = \int_{t_0}^{t_1}\sqrt{g_{ab}\frac{dx^a}{dt}\frac{dx^b}{dt}}\,dt $$ but if we choose $t=s$ we have $$ L = \int_{s_0}^{s_1} ds = s_1-s_0\,. $$ This is called "proper time" or sometimes "unit speed". This does not help compute the length of the curve however, since we still need to evaluate $s_1-s_0$.

The geodesic equations arise from minimizing an action. There are two equivalent choices, one can minimize the length $L$ or one can also minimize the energy functional $$ I = \int_{t_0}^{t_1}g_{ab}\frac{dx^a}{dt}\frac{dx^b}{dt}\,dt $$ The actions $L$ and $I$ are different but their variations both lead to the geodesic equations.

Not sure I answered exactly the question, the text your using sounds a little imprecise (maybe try Sean Carroll's book Spacetime and Geometry), but I hope it helps.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.